File size: 5,264 Bytes
9dbe682 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 | import gradio as gr
from transformers import AutoTokenizer, DebertaV2Config, DebertaV2Model, PreTrainedModel
import torch
import torch.nn as nn
import torch.nn.functional as F
import re
# ==============================================================================
# SECTION 1: TEXT CLEANING
# ==============================================================================
def clean_teks(text):
text = text.lower()
text = re.sub(r"http\S+|www\S+|https\S+", "", text)
text = re.sub(r"[^a-zA-Z0-9\s]", "", text)
text = re.sub(r"\s+", " ", text).strip()
return text
# ==============================================================================
# SECTION 2: MODEL DEFINITION
# ==============================================================================
class DebertaV3ForMultiTask(PreTrainedModel):
config_class = DebertaV2Config
def __init__(self, config):
super().__init__(config)
self.num_sentiment_labels = config.num_sentiment_labels
self.num_type_labels = config.num_type_labels
self.deberta = DebertaV2Model(config)
self.dropout = nn.Dropout(config.hidden_dropout_prob)
self.sentiment_classifier = nn.Linear(config.hidden_size, self.num_sentiment_labels)
self.type_classifier = nn.Linear(config.hidden_size, self.num_type_labels)
self.init_weights()
def forward(self, input_ids=None, attention_mask=None, **kwargs):
outputs = self.deberta(input_ids=input_ids, attention_mask=attention_mask)
hidden_state = outputs.last_hidden_state[:, 0]
pooled_output = self.dropout(hidden_state)
sentiment_logits = self.sentiment_classifier(pooled_output)
type_logits = self.type_classifier(pooled_output)
return {
"sentiment": sentiment_logits,
"type": type_logits,
}
# ==============================================================================
# SECTION 3: LOAD MODEL & TOKENIZER
# ==============================================================================
MODEL_PATH = "./finetuned_model_deberta_multitask"
tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
model = DebertaV3ForMultiTask.from_pretrained(MODEL_PATH)
model.eval()
SENTIMENT_LABELS = ['negative', 'neutral', 'positive']
CATEGORY_LABELS = ['Business', 'Entertainment', 'General', 'Health', 'Science', 'Sports', 'Technology']
# ==============================================================================
# SECTION 4: PREDICTION FUNCTION
# ==============================================================================
def predict(text):
if not text or text.isspace():
return {}, {}, "No input provided", "No input provided"
cleaned_text = clean_teks(text)
inputs = tokenizer(cleaned_text, return_tensors="pt", truncation=True, max_length=256, padding=True)
with torch.no_grad():
outputs = model(**inputs)
sentiment_logits = outputs["sentiment"]
type_logits = outputs["type"]
sentiment_probs = F.softmax(sentiment_logits, dim=1)[0]
type_probs = F.softmax(type_logits, dim=1)[0]
sentiment_confidences = {label: round(prob.item(), 4) for label, prob in zip(SENTIMENT_LABELS, sentiment_probs)}
category_confidences = {label: round(prob.item(), 4) for label, prob in zip(CATEGORY_LABELS, type_probs)}
best_sentiment = SENTIMENT_LABELS[torch.argmax(sentiment_probs)]
best_category = CATEGORY_LABELS[torch.argmax(type_probs)]
return sentiment_confidences, category_confidences, f"{best_sentiment} ({sentiment_confidences[best_sentiment]:.2%})", f"{best_category} ({category_confidences[best_category]:.2%})"
# ==============================================================================
# SECTION 5: GRADIO UI
# ==============================================================================
with gr.Blocks(theme=gr.themes.Soft()) as demo:
gr.Markdown("## 📰 News Sentiment and Category Classification")
text_input = gr.Textbox(placeholder="Enter news text here...", label="Input Text", lines=5)
submit_button = gr.Button("Analyze", variant="primary")
with gr.Row():
with gr.Column():
gr.Markdown("### 🔎 Predicted Sentiment")
sentiment_label = gr.Text(label="Predicted Sentiment")
sentiment_output = gr.Label(label="Sentiment Probabilities", num_top_classes=3)
with gr.Column():
gr.Markdown("### 🗂️ Predicted News Category")
category_label = gr.Text(label="Predicted Category")
category_output = gr.Label(label="Category Probabilities", num_top_classes=len(CATEGORY_LABELS))
submit_button.click(fn=predict, inputs=text_input, outputs=[sentiment_output, category_output, sentiment_label, category_label])
gr.Examples(
[
["Stanley Kubrick's estate has led the tributes to Shelley Duvall."],
["Lignetics Inc. recently acquired the fiber energy products wood pellets business unit from Revelyst."],
["An overcrowded California men’s prison was running on emergency generator power for a third day Tuesday."]
],
inputs=text_input
)
if __name__ == "__main__":
demo.launch() # No need for share=True when deploying
|